Published 1 November 2026 · Text V1 · Agentic package A1

2: LLMs as Pure Epistemic Technologies

This chapter examines what distinguishes large language models from earlier epistemic technologies in the social sciences and humanities. Building on Alvarado’s account of epistemic technology, it argues that LLMs simulate inquiry by directly producing argumentative texts that take the form of its completed results. This capacity motivates the concept of pure epistemic technology for disciplines in which knowledge is produced and communicated through argumentative discourse. The chapter then connects this account to the extended mind: LLMs can generate formulations, conceptual connections, and argumentative structures that were unavailable to the researcher in that form before the interaction. Their integration extends epistemic functions such as conceptual articulation and synthesis. Decisions about delegation, dialogue, iteration, and validation therefore shape the cognitive architecture of the researcher–LLM system. Different distributions of these functions have different epistemic consequences. The chapter identifies this generative extension and the resulting configuration choices as problems requiring an epistemological assessment of how researchers integrate LLMs into inquiry.

Release information

Citation

Frega, Roberto. 2026. “2: LLMs as Pure Epistemic Technologies.” In The Driver’s Seat: An Introduction to the Epistemology of LLMs. Version V1 / A1. Zenodo. https://doi.org/10.5281/zenodo.23856605.